18 of the 19 US high-net-worth investors with $2M or more chose a named advisor over an instant AI assistant, and 1 of 30 would move money for better AI
Forced to choose one, 18 of the 19 US high-net-worth clones holding $2m or more took a named advisor they could reach in two days over an AI assistant available instantly, while 4 of the 7 below $1m took the AI (n=30, September 2026).
- Who was asked
- Brox HNWI, 30 clones, US
- When
- September 2026 (fielded 15 September 2026)
- How
- Predicted responses from Brox clones, each built from a long-form interview with a real person.
- What changed from the obvious question
- The obvious question, "would you use an AI assistant from your bank?", collects an attitude. Instead the panel was made to give something up: one forced choice between a named advisor reachable in two business days and an instant assistant, then seven concrete situations one at a time, so the answer is a handover point rather than an opinion about AI.
Section 01: What the panel said
- #
18
Forced to choose one, 24 of 30 US high-net-worth investors took a named advisor they could reach within two business days over an AI assistant available instantly at any hour, and appetite runs the wrong way up the wealth curve: 18 of the 19 holding $2m or more took the advisor, while 4 of the 7 below $1m took the AI.
base 19 · $2m or more in household liquid assets
- #
23
They will let AI near the advisor and not near the advice: 23 of 30 are comfortable with an advisor using AI to prepare meetings and take notes, yet 15 of 30 say nothing would make them act on AI advice alone, and 4 of 30 would act because it showed them its reasoning.
base 30 · contrast
- #
1
AI does not win the account: asked what would most make them move money elsewhere, 15 of 30 named a better advisor and 1 of 30 named better digital tools and AI, and 29 of 30 want to be told, always, when an answer came from AI.
base 30 · caveat
- #
30
The handover point is the size of the decision, not the difficulty of the question: all 30 want a person for an inheritance or a large family gift and 29 of 30 for tax planning before year end, against 10 of 30 for a routine statement question.
base 30 · contrast
The handover point is the size of the decision, not the difficulty of the question: all 30 want a person for an inheritance or a large family gift and 29 of 30 for tax planning before year end, against 10 of 30 for a routine statement question.
02Who do wealthy clients want on the other end?
We gave 30 US high-net-worth clones one forced choice. Either a named human advisor they can reach within two business days, or an AI assistant available instantly, at any hour, that answers most questions well. 24 of 30 took the advisor.
What makes it interesting is where the six who took the AI sit. The panel splits into four wealth bands: 7 clones with $500,000 to $1m, 4 with $1m to $2m, 8 with $2m to $5m, and 11 with $5m or more. The advisor was chosen by 3 of 7 in the lowest band, 3 of 4 in the next, 8 of 8 in the third and 10 of 11 in the top. So 18 of the 19 clones holding $2m or more took the person, and below $1m it inverts: 4 of the 7 took the machine. Appetite for an always-on assistant runs the wrong way up the wealth curve, which is the opposite of how these things are usually rolled out.
03What will they actually let AI do?
A good deal, as long as it stays behind the advisor. 23 of 30 are comfortable with an advisor using AI to prepare for their meetings and take notes during them. Then it stops. 15 of 30 say nothing would make them act on AI advice alone; 15 of 30 would act if a human advisor had reviewed it first; and 4 of 30 would act because it showed them the reasoning behind the recommendation.
The handover point is the size of the decision rather than the difficulty of the question. Asked who they would want handling each of seven situations, all 30 wanted a person for a decision about an inheritance or a large family gift, 29 of 30 for tax planning before year end, 28 of 30 for a 10% market drop in a single week, 28 of 30 for an unexpected fee and 27 of 30 for rebalancing the portfolio. Only two situations fall away: idle cash or a maturing certificate of deposit, at 18 of 30, and a routine question about a statement, at 10 of 30.
04Does better AI move any money?
Barely. Asked which one thing would most make them move money to a different institution, 15 of 30 named a better advisor, 7 of 30 better investment returns, 5 of 30 lower fees, 2 of 30 said nothing would move them, and 1 of 30 named better digital tools and AI. On the other side of the same question, 18 of 30 said AI involvement would make no difference at all to their confidence in a recommendation, 8 of 30 said it would lower it and 4 of 30 said it would raise it.
Disclosure is the closest thing to a unanimous answer in the study: 29 of 30 want to be told, always, when an answer came from AI rather than from a person. Asked how they would want that done, 17 of 30 chose a human-review stamp on the answer and 12 of 30 chose a confidence score with the rationale. And when an assistant gets something wrong and it costs them money, 17 of 30 would hold themselves responsible for acting on it, 12 of 30 the bank and 1 of 30 their advisor personally.
05What would a wealth platform team do with this?
- Our reading, not the panel’s: launch to the affluent tier before the top of the book. The appetite is at the bottom of the wealth curve, where 4 of 7 took the machine, and carrying the same experience upmarket unchanged is where it stops working.
- Build it behind the advisor first. 23 of 30 are comfortable with preparation and note-taking, which is the part with no consent problem and most of the time saved.
- Ship the routine questions and stop. Statements and idle cash are the only two situations where a majority or near-majority will take a machine. Everything else, up to and including an unexpected fee, is a person’s job in the client’s mind.
- Treat disclosure as a hygiene factor, not a feature. 29 of 30 want telling every time, and 17 of 30 want the human-review stamp specifically, which means the review has to be real before the stamp can be shown.
- Do not expect AI to win an account. 1 of 30 would move money for it, and 15 of 30 would move for a better advisor. It is a reason to stay rather than a reason to switch.
Section 06: What they said, and the interview line it rests on
“I would definitely choose a named human advisor, even if it means waiting a day or two to reach them. I prefer person-to-person communication because it is the most efficient way of solving problems. Everything done electronically just feels like a follow-up.”
Predicted from this line in the interview
“I prefer person-to-person communication because this is the most efficient way of solving problems. Everything done electronically is a follow-up.”
Section 07: How much should you trust this?
What we would defend
- The direction up the wealth curve: 18 of the 19 clones at $2m or more chose the named advisor, against 3 of the 7 below $1m, on bands of 7, 4, 8 and 11.
- The split between preparation and delivery: 23 of 30 comfortable with an advisor using AI to prepare, 15 of 30 who would act on AI advice alone under no circumstances.
- The ranking of situations, which moves in one direction from 30 of 30 wanting a person for an inheritance down to 10 of 30 for a routine statement question.
- That disclosure is close to unanimous: 29 of 30 want telling, always, when an answer came from AI.
What we would not
- Any of these as a measurement. Thirty clones, and the wealth bands behind the split are 7, 4, 8 and 11 people, so every figure is a count and the finding is the direction rather than the level.
- This as anybody's own client book. It is the Brox HNWI panel, so it says what wealthy Americans expect of a private bank in general, not what one firm's clients think of that firm.
- The levels of resistance. Nobody was shown a working assistant, and stated resistance to a category usually softens once the thing is in front of people, so read the levels as conservative and the ranking as the finding.
The main limitation
Thirty clones. Every figure here is a count, and the wealth bands behind the split are 7, 4, 8 and 11 people. Read it as direction, not as a measurement.
Section 08: How was this run?
| Segment | Clones |
|---|---|
| $500k to $1m | 7 |
| $1m to $2m | 4 |
| $2m to $5m | 8 |
| $5m or more | 11 |
Question wording as fielded
- For each of the following, who would you want handling it? — A cash balance or maturing CD you should probably move
- For each of the following, who would you want handling it? — A market drop of 10% in a single week
- For each of the following, who would you want handling it? — Rebalancing your portfolio
- For each of the following, who would you want handling it? — Explaining a fee you did not expect
- For each of the following, who would you want handling it? — A decision about an inheritance or a large family gift
- For each of the following, who would you want handling it? — Tax planning before year end
- For each of the following, who would you want handling it? — A routine question about your statement
- Would you want to be told when an answer came from AI rather than from a person?
- Which statement is closest to how you feel about contact from the institution that looks after most of your money?
- An AI assistant from your bank recommends a change to your money. What would make you act on it?
- Your bank's AI assistant gets something wrong and it costs you money. Who do you hold responsible?
- Roughly how much of your total investable wealth sits with the single institution you consider your main one?
- Your bank offers an always-on AI assistant that can answer questions about your portfolio, flag opportunities and risks, and let you act on them straight away. How interested would you be in using it? 0 means not at all interested, 10 means extremely interested.
- If you had to choose one, which would you rather have from your bank?
- How would you feel about your adviser using AI to prepare for your meetings and to take notes during them? 0 means very uncomfortable, 10 means very comfortable.
- Which one of these would most make you move money to a different institution?
- What would be the biggest concern that would stop you from trusting an AI assistant with your portfolio decisions?
- How would you most want your bank to disclose AI involvement when you get an answer or recommendation?
- What is the one question about your own money that you most wish someone would just answer clearly?
- If your bank's AI assistant meant you spoke to a human adviser less often but got answers much faster, would the relationship be better or worse? 0 means much worse, 10 means much better.
- If an AI assistant gives you guidance about your money, when would you want a human adviser to review it?
- When you think about keeping more of your money with your primary institution, which digital or AI capability would make the biggest difference to you?
- Thinking about advice on your own money, what is the main reason you would want a human adviser involved instead of an AI assistant?
- For each of the following situations, how much explanation would you want before you would act on AI guidance? — A recommendation to rebalance your portfolio
- For each of the following situations, how much explanation would you want before you would act on AI guidance? — A warning that your portfolio risk has increased
- For each of the following situations, how much explanation would you want before you would act on AI guidance? — A suggestion to move idle cash
- For each of the following situations, how much explanation would you want before you would act on AI guidance? — A prompt to realize gains or losses before year end
- For each of the following situations, how much explanation would you want before you would act on AI guidance? — A recommendation tied to a major market move
- Which of these best describes how AI involvement would affect your confidence in a recommendation from your primary wealth institution?
- When it comes to your own money, how much of the question 'Am I financially okay?' would you be comfortable having answered by an AI assistant from your bank rather than by a person? 0 means none of it, 10 means all of it.
What was not asked, and why
Nobody was shown a working assistant, or any interface at all, so this is what people expect of the idea rather than how they behave in front of one. No fee was attached to any option, so the trade between an advisor and an assistant was never priced. The panel is the Brox HNWI panel rather than any one institution's clients. Ask Brox added seven questions of its own to the thirteen written for the study, and all twenty are in the data table.
What changed from the obvious question
The obvious question, "would you use an AI assistant from your bank?", collects an attitude. Instead the panel was made to give something up: one forced choice between a named advisor reachable in two business days and an instant assistant, then seven concrete situations one at a time, so the answer is a handover point rather than an opinion about AI.
More on AI in investing and due diligence · Wealth & HNWI · Brox HNWI (US) panel
Ask this of your own clients: humans@brox.ai
Related insights
All in this topic-
Angel and VC investors drop an AI startup when its margin after AI model costs falls from 70 to 45 cents on the dollar: 12 of 30 would likely invest at 70 cents, 2 at 45, and none at 20
12 Offered $50,000 of an AI startup selling software to businesses, valued at $20 million with $1 million a year in revenue growing three times a year, 12 of the 30 private-market investors rated themselves likely to invest (4 or 5 on a five-point scale) when it keeps 70 cents of every revenue dollar after paying for its AI models, 2 of 30 at 45 cents and 0 of 30 at 20 cents; no clone rated a lower margin higher than a higher one (base 30)
Read -
Angel and VC investors turned down an AI-run venture fund at every fee tested: 1 of 30 chose it at 2% and 20%, and still only 1 at a quarter of that
1 A venture fund in which an AI picks every startup, with no human able to overrule it, drew 1 of the 30 private-market investors at equal fees of 2% a year plus 20% of profits, and the same 1 at 1% and 10% and at 0.5% and 5%, while a human-run fund at 2% and 20% drew 14, 10 and 11; of the 29 who passed on the AI fund at equal fees, 0 chose it at either lower fee (base 30)
Read